Clinical Characteristics and Hypokalemia Risk Prediction in Patients With Hyperosmolar Hyperglycemic State and Diabetic Ketoacidosis (HHS-DKA): A Single-Center, Retrospective Observational Study.

Background: People with combined diabetic ketoacidosis (DKA) and hyperosmolar hyperglycemic state (HHS) often present with more severe metabolic derangements than those with DKA or HHS alone. This study aimed to clarify the clinical characteristics of HHS-DKA and explore predictive models for complications, including hypokalemia.

Methods:  We retrospectively analyzed data from 99 patients admitted with hyperglycemic emergencies between April 1, 2010, and October 31, 2024, and classified them into DKA, HHS, and HHS-DKA groups. A decision tree model was also developed to predict the risk of post-continuous insulin infusion (CII) hypokalemia. The decision tree model was created using machine learning with the Python language (Python Software Foundation, Wilmington, Delaware).

Results: HHS-DKA patients had significantly higher rates of acute kidney injury (84%) and hyperkalemia (58%) compared to those with DKA or HHS alone. A decision tree model predicted post-CII hypokalemia with 80% accuracy, identifying key predictors such as initial blood glucose and insulin flow rates.

Conclusion: HHS-DKA represents a distinct and severe clinical entity with unique characteristics and complications. Predictive models developed in this study will likely assist in risk stratification and improve patient management during hyperglycemic crises in emergency settings. However, as this was a single-center retrospective study without external validation, further studies are warranted to confirm these findings.